ProFAD: Probabilistic Falsification for Mapping Unsafe Boundaries in Autonomous Driving

This project (ProFAD) will develop a simulation-based framework that moves beyond collecting isolated failure cases, toward systematically mapping the boundaries of unsafe operation with probabilistic guarantees. Current stress testing and adversarial methods can surface rare crashes, but they typically return only a handful of failing scenarios and do not characterize how large the unsafe region is, how it evolves under noise and domain shifts, or whether nearby conditions share the same vulnerability. ProFAD addresses this gap by combining formal, structured scenario specifications (e.g., Scenic) with an adaptive, partition-based falsification engine that concentrates sampling near suspected safe/unsafe boundaries and uses surrogate models to improve sample efficiency while maintaining calibrated confidence statements. A key component is the use of Generative Flow Networks (GFlowNets) to guide diverse, reward-driven sampling of rare but safety-critical scenarios: rather than optimizing for a single “worst case”, GFlowNets enable ProFAD to discover and maintain multiple distinct failure modes and to cover the unsafe boundary more comprehensively. The end product is a set of certified unsafe regions (not just traces), integrated with high-fidelity simulators such as CARLA/AWSIM, producing interpretable “vulnerability maps” that support reproducible, scalable safety assessment of autonomous driving systems.

• Record URL:
  • https://www.clemson.edu/cecas/tracr/research/projects/#profadprobabilisticfalsificationformappingunsafeboundariesinautonomousdriving

Language

• English

Project

• Status: Active
• Funding: $278,478.00
• Contract Numbers:

  69A3552344812

  69A3552348317

• Sponsor Organizations:

  Office of the Assistant Secretary for Research and Technology

  University Transportation Centers Program
  Department of Transportation
  Washington, DC  United States  20590

  Purdue University

  1040 South River Road
  West Lafayette, IN  United States  47907

  University of California, Santa Cruz

  1156 High Street, Mail Stop SOE2
  Santa Cruz, California  United Kingdom  95064

  Morgan State University

  1700 E. Coldspring Lane
  Baltimore, Maryland  United States  21251
• Managing Organizations:

  National Center for Transportation Cybersecurity and Resiliency (TraCR)

  Clemson University
  Clemson, SC  United States 

  Purdue University

  1040 South River Road
  West Lafayette, IN  United States  47907
• Project Managers:

  Chowdhury, Mashrur

• Performing Organizations:

  Purdue University

  1040 South River Road
  West Lafayette, IN  United States  47907

  University of California, Santa Cruz

  1156 High Street, Mail Stop SOE2
  Santa Cruz, California  United Kingdom  95064

  Morgan State University

  Department of Transportation and Urban Infrastructure Studies
  1700 E. Cold Springs Lane
  Baltimore, MD  United States  21251
• Principal Investigators:

  Ukkusuri, Satish

  Cardenas, Alvaro

  Fremont, Daniel

  Celik, Berkay

  Ali, Amjad

• Start Date: 20260401
• Expected Completion Date: 20270331
• Actual Completion Date: 0
• USDOT Program: University Transportation Centers

Subject/Index Terms

• TRT Terms: Autonomous vehicles; Simulation; Vehicle operations; Vehicle safety
• Subject Areas: Data and Information Technology; Highways; Safety and Human Factors; Vehicles and Equipment;

Filing Info

• Accession Number: 01988089
• Record Type: Research project
• Source Agency: National Center for Transportation Cybersecurity and Resiliency (TraCR)
• Contract Numbers: 69A3552344812, 69A3552348317
• Files: UTC, RIP
• Created Date: Apr 28 2026 3:40PM